[cclib-devel] Test files

[Noel O'B. <no...@ca...>]

I've just been doing some tidying up of the tests. I've removed the
so-called wild folders, and used our new Googlemail system instead. Here
are some proposed guidelines for dealing with new problem output files:

(0) Give it a unique name (compared to others in the repository)
(1) gzip it
(2) Email it to ccl...@go... with the following info:
(a) What's the problem? "Breaks parser" is fine, if that's all that we
know already.
(b) What revision does it have a problem with.
(c) Where does the file come from, and confirm that it's public domain
[I think we only spoke briefly about this before, but I believe that we
need all test log files to be public domain...are you happy with this
for your log files?? I have already asked Joe Townsend, and the AOMIX
files are, by nature, public domain]
(d) Where will the file be stored in the repository (not sure is this
necessary)
(3) The log message of the revision that fixes the problem should refer
to the name of the problem log file (as well as the usual info).
(4) Login to googlemail and reply to the original problem with the name
of the revision that fixes it. This lets us know at a glance (on
googlemail) whether a particular log file is still a problem. (Although
we should also know this from regression.py)

Somewhere in all of this, preferably after (2), to add a regression test
to regression.py for the problem. This will let the other developer know
that there's a problem that needs to be fixed and also that they are
missing a test file in their local repository. I have already added some
broken parser tests for the 4 files in the test repository (haven't done
anything about that second ADF example of yours).

I will try to set up an automated procedure to make a .zip of all of the
data files in my local repository (excluding the basic ones), and send
it up to sourceforge. I think there are ways to do this, but it will
take some time.

I'll wikify these instructions if they sound reasonable - please let me
know any further ideas.

Oh yeah, could you send to google that Gaussian example you said there
was a problem with?

Regards,
 Noel